Transcript crooks

Tera-Pixel APS for CALICE
Progress meeting, 17th September 2007
Jamie Crooks, Microelectronics/RAL
Laser Testing
• Proof of life from both
test pixels
• No obvious injection
from laser itself
Comparator & Trim
•
•
Comparator trim biases
adjusted to give sensible
step size
Holding test pixel in reset
(ie zero input), record
switching threshold for
each trim setting
Upper and lower limit of
switching point recorded:
Comparator threshold adjustment
160
lo Trimbias = 2048
lo Trimbias = 4000
hi Trimbias = 4000
120
100
NB: One DAC step ~0.4mV
Comparator noise
6
80
60
Trimbias = 2048
Threshold Range, DAC (mV)
hi Trimbias = 2048
140
DAC (mV)
•
Trimbias = 4000
5
40
4
3
20
2
1
0
0
0
0
2
4
6
8
10
TRIM (4bit setting)
12
14
16
2
4
6
8
10
TRIM (4bit setting)
12
14
16
Digital Logic
• SRAM problem reported last week is now understood
– Sense biases adjusted (again) to properly operate sense amplifiers (read)
– SRAM write is reliable at VDD2V5dig > 2.6v
• 20mA -- 40mA static current in digital ground on-chip
• Noted useful characteristic
– SRAM write of 1s requires 2.6v, write of 0s requires ~2.3v
• Workaround possibilities:
Run the sensor at 2.6v
– Does work
• No software yet to
evaluate performance
– Not ideal
• Evidence that this
raises the internal
digital ground
“Reset” the SRAMs at 2.6v then switch to lower
supply for normal operation
– Can use the spare DAC channel and transistor to
supply current: Board mod TBC.
– Slow (software controlled DAC serial write) but only
required once between each bunch train
– Requires longer in-spill timing (separation of MUX &
clock signals) TBC
Digital Activity
One column
of data
•
Override mode
generates a full
memory of hits
(to be understood  next slide)
One row
of data
One bunch crossing of data
Digital: Odd behaviour
Column 1
Column 2
Column 3
Column 4
(preShape)
(preShape)
(preSample)
(preSample)
Override mode
generates a full
memory of hits
First 6 registers
in each row
incorrectly read
000000 in the hit
field!?
Third timecode: 5 registers
used (no more available)
•
•
Second timecode: 7 registers used
First timecode: 7 registers used
Not yet understood – suspect some internal timing/propagation delay might be
affecting memory controller
Could workaround by effectively disabling those registers (would reduce memory
capacity in those columns)
Testing Status
•
Sensor basic functionality: working!
– Known design errors with workarounds:
1
Current limiting inverter
(monostables)
Remove transistor, add resistor to
ground on PCB
2
Sense amplifier bias chain
Resistor value change on PCB
3
SRAM write (3.3v not possible)
Run sensor at 2.6v, or
SRAM reset, variable power supply and
change to in-spill timing
– One remaining “feature” to be investigated in the digital logic (SRAM)
• (Some additional firmware/software features to aid diagnosis)
•
Sensor design evaluation
– Now need higher level software to facilitate
• Noise rate/distribution
• Threshold sweeps etc
• Per [row/col/pixel] mask and trim verification
– Analog performance
• New boards & system for Giulio & Marcel
Additional CALICE test structures
CfI grant “IMAPS” for development/exploration of sensors on INMAPS
process (this FY) – some possible spare silicon available for us…
•
preShape pixel (in its current form) with analog outputs
– Not time to include on original design
– Will further our understanding of analog performance of this process
– Very little design effort required
•
With
neighbouring
inactive pixels
preSample pixel (in its current form) with analog outputs
– Simply for reference/comparison
•
Bump bond trial pads
– Second CALICE ASIC will need to be tiled to make a 15x15cm sensing area for
beam test
– Most area efficient way to to this will be to bump bond the chips
– Solder bump preferred solution
– Some large pads suitable for bump bonding trials ahead of the second CALICE
sensor
– Small design effort
Would be sent on multi-project run submission mid Jan 2008